1

Shareholder Protection and the Cost of Capital:Empirical Evidence from German and Italian Firms

Julie Ann ElstonUniversity of Central Florida

Laura RondiCERIS-CNR

April 19, 2004

Abstract

We investigate the cost of capital in a model with agency conflicts between inside and outsideshareholders, where the severity of agency costs depends on a parameter representinginvestor protection that is expected to vary between countries with different degrees ofshareholder protection. Using recent firm-level data from two countries with low shareholderprotection, Italy and Germany, we estimate the predicted relationships among investorprotection, inside ownership, and the marginal cost of capital for both old-economy andyoung new-economy firms. Results indicate that in Italy, a high concentration of insiderownership results in a higher cost of capital. While R&D intensive firms in both countriesare more likely to have strong insider ownership.

JEL Classification: G31; G32; E22; D92; O16

Key Words: Shareholder protection, ownership structure, cost of capital, agency costs

Julie A. Elston: Department of Economics, CBA, PO Box 161400, University of Central Florida,Orlando FL, 32816-1400, USA. Phone: +1(407)8232078. Email: julie.elston@bus.ucf.edu

Laura Rondi: CERIS-CNR, Institute for Economic Research on Firms and Growth, Via Avogadro 8,10121 Torino, Italy. Phone: +39-011-5601209, Fax: +39-011-5626058. L.Rondi@ceris.to.cnr.it

2

1. Introduction and motivation

This paper explores the consequences of poor investor protection for the growth of firms in

Europe. Over the past decade, European policy-makers’ concerns about the slow growth of

employment, income, capital formation, entrepreneurship and high-tech investment have motivated

new policy initiatives to promote the development of equity markets, foster financial markets

integration and improve legal harmonization. A key rationale behind the EU public policy is that,

similar to the US, it believes that by providing firms with better access to equity finance and

venture capital, they would grow more rapidly than internal finance would allow.1 However in spite

of the development of European stock markets in the 1990s, economic growth in the EU remains

weak and recent evidence at the firm-level shows that even when companies “go public”, their

subsequent growth rates remain low, especially when compared to US firms.2

If the growth incentives are a dominant issue for US firms going public, but not for European

firms, this hints at differential risk aversion or cost of capital between EU and the US, assuming

investment opportunities are equally distributed across countries. Differences in the financial

contracting environment are emphasized by broad international studies within the Law and Finance

literature, indicating that differences in the way legal systems protect investors may ultimately help

to explain differential growth rates (Demirguc-Kunt and Maksimovic, 1998, Levine, 1999).

Recent studies by La Porta, Lopez-de-Silanes, Shleifer, and Vishny (“LLSV”, 1997, 1998,

2000), have shown that investor protection may be important in explaining why firms are owned

and financed very differently in between countries. For example, cross-country evidence shows that

firms in countries with weak investor protection, such as many EU member states, have highly

concentrated ownership (LLSV, 1999 and Barca and Becht, 2001). Himmelberg, Hubbard, and

Love (2002) expanded this work by deriving a structural econometric model in which the effects of

the legal system are summarized to investigate the relationship among investor protection, inside

ownership concentration, and the cost of capital -where the empirical specification allows for firm-

level dimensions of investor protection. The model is important because it provides a framework to

explain the role of weak governance in the efficient accumulation of capital.

In this paper, we empirically investigate the impact of investor protection on the cost of

capital in Europe, where investor protection characterizes what HHL (2002) collectively refer to as

1 A large body of recent empirical literature has examined the impact of financing constraints on firms’ fixed capitalinvestment decision (see Hubbard, 1998, for a survey).

3

“those features of the legal, institutional, and regulatory environment –and characteristics of firms

or projects- that facilitate financial contracting between inside owners (managers) and outside

investors” (p. 2).

Our study extends previous work examining differences across countries by 1) providing a

micro examination of firm behavior and agency related problems for young firms in the 1990s in

Germany and Italy, and 2) allowing comparison of two allegedly similar poor shareholder

protection countries which in fact have two very different institutional environments.

In countries where investor protection is said to be generally low, the HHL model predicts

endogenously high levels of insider ownership. Accordingly, the idiosyncratic risk premium

applied to the cost of capital should be high, implying a steady-state level of capital below the first-

best level. This allows us to measure the real effects of corporate governance, namely the effects on

the accumulation of capital. In this paper we estimate the determinants of the fraction of equity

owned by insiders, testing, as predicted, whether this fraction depends on measures of investor

protection. Further we test the HHL prediction that investor protection has an important cross-firm

and cross-cultural dimension. In addition, we investigate the correlation between inside equity

ownership and the marginal return to capital, a relationship that follows directly from the first-order

condition for capital –the cost of capital includes a risk premium that reflects the insiders’ exposure

to idiosyncratic risk.

This paper utilizes two new and unique data sets tracking the behavior of Italian and German

firms listed on the Borsa Italiana and the Neuer Markt during the 1990s. In addition to the financial

data from annual reports of these firms, our study also uses information on corporate ownership

structure, identity of investors, inside ownership, and characteristics of the IPO.

The paper is organized as follows. Section 2 illustrates the theoretical framework and the

empirical implications of our model. Section 3 briefly describes the institutional context in

Germany and Italy, and the data. Section 4 describes the empirical strategy and presents the

results. We conclude in Section 5 by discussing implications for policy makers concerned with

addressing small firm growth and financing issues within the European context.

2 See Pagano, Panetta and Zingales (1998) and, for more recent evidence, Carpenter and Rondi (2003).

4

2. The theoretical framework

The theoretical foundation of this study is based on the agency theory of the firm as outlined

in Alchian and Demsetz (1972) and Jensen & Meckling (1976), where agency problems between

insiders and outsiders can arise because insiders can divert firm profits to themselves before paying

dividends. There is also a relevant treatment of these issues in the Law and Finance literature,

where different legal institutions reportedly have systematic influences on investor protection and

ownership structure that differ across countries. More recently, the HHL (2002) model blends

these two strands of literature, introducing insider risk aversion as the offsetting cost of insider

ownership.

Apparently, poor investor protection (or more pointedly, poor protection of minority

shareholders) favors insiders (controlling shareholders who control decision-making) because it

allows them to expropriate outside (minority) shareholders. Concentrated ownership is thus viewed

as the response to the lack of legal protection that reduces agency problems and managerial slack

(Jensen and Meckling, 1976; Shleifer and Vishny, 1986), or minority investors’ expropriation -

under the assumption that the dominant shareholder does not steal from oneself.3

We argue that the full consequences of lack of investor protection are not fully understood,

particularly the implications for the firms’ growth. Weak (minority) investor protection and

concentrated ownership may be detrimental to both the insider and the firm in that they prevent

insiders from diversifying risk optimally and the firm from pursuing growth-oriented, risky capital

projects, thus impeding optimal capital accumulation. The purpose of this study is to empirically

examine these issues using data from two countries with different environments yet both weak

shareholder protections in order to comment on similarities and differences within this group.

Under imperfect investor protection, the entrepreneur/manager has to retain an equity stake in

the firm large enough to reassure minority shareholders that he will neither pursue value-destroying

projects nor carry-out expropriation. The insider will thus be forced to bear high levels of

idiosyncratic risk for having a large bulk of his wealth invested in (tied to) the firm. By reflection,

the insider will have to raise external funds (equity) in proportion to the initial wealth invested in

3 Interestingly, reliance on this “second-best” solution to agency frictions, however, may be the reason why, in manyindustrialized countries, company by-laws still include relatively few rules aimed at protecting minority shareholders.Although some progress has been achieved within the EU, as part of the legal harmonization program, but much is yetto be done with full harmonization is far from being accomplished. For example, recent financial scandals in Italy (e.g.Parmalat and Cirio) have been interpreted as a consequence of the delay in the harmonization (See The Economist,January 3rd, 2004).

5

the firm that does not dilute his incentives, independent of the actual amount of equity finance that

would be needed to fund the firm’s growth project.

We expect two consequences from this. First there should be an additional premium in the

cost of capital that reflects the additional idiosyncratic risk that the entrepreneur/manager is forced

to bear, thus reducing in equilibrium, the desired capital stock. And second, a suboptimal quantity

of capital is raised on the equity market (at the time of the IPO, and subsequently).

2.1 Insider ownership, stealing, and the cost of capital

The HHL (2002) model can be summarized as follows. The model describes a firm where

managers are in full control of the decision-making, and have access to a growth project. The firm

has a Cobb Douglas production function, described by Π(K), where K is the capital stock that

depreciates at the rate δ. The managers are risk averse and seek to diversify by selling a fraction 1-

α, of the equity in the firm. Insiders can steal or divert a fraction sit+1 of firm profits to themselves

before paying dividends, but stealing has a cost, as defined by an exogenous punishment

technology, which is a positive function of a quantitative index of investor protection, φit4 :

c(φit, sit) = ½ φitsit2 [1]

The manager’s net return Nit+1 in period t+1, after taking the firm public, is:

Nit+1 = [αit(1 - sit+1 ) + sit+1 - c(φit, sit+1)]Π(K it+1,θ it+1) [2]

Because of agency problems between insiders and outsiders, when managers have to raise

external finance they have to convince outside investors that they will receive a fair market rate of

return (i.e. that stealing will not occur).5 With imperfect investor protection, the managers have to

commit to lower levels of future stealing by retaining a higher fraction of equity than would be

optimal for them to fully diversify the firm-specific risk. Consequently, they are forced to bear high

4 Higher values of the parameter φit indicate better protection.5 The level of stealing that maximizes the managers’ net return (∂Nit+1 / ∂sit+1) is characterized by the first-ordercondition cs(φit, sit+1) + αit = 1. The optimal stealing is therefore: sit+1 = (1-αit)/φit which indicates that stealing isincreasing in outside ownership, decreasing in investor protection, and also that inside ownership is inversely related toinvestor protection.

6

levels of diversifiable idiosyncratic risk. The tradeoff between risk and insiders’ incentive to invest

in risky capital projects determines the inside ownership stake in equilibrium. The empirical

implications derive from the first-order condition that characterizes the optimal capital choice:

E [Π’(k)] = r + δ + Γ + α(γ – Γ) [3]

E [Π’(k)] is the marginal profit of capital and the right hand side represents the firm’s user

cost of capital, where r is the risk-free rate and δ is the depreciation rate on capital. Γ and (γ – Γ)

represent the risk discounts in the cost of capital for (non diversifiable) systematic risk and

(diversifiable) idiosyncratic risk, which depend in turn on the stochastic discount factors of the

market and of the manager, respectively. What is of interest here is that the idiosyncratic

component exists because a large fraction of the insider’s income is derived from the profitability

of the firm. With poor investor protection, and α>0, there is an additional premium in the cost of

capital, namely, α(γ – Γ) > 0.

The economic intuition of the model is that insiders assign a lower value to risky projects

(and profits) than outside investors. Assuming (r + δ + Γ) constant, one can estimate (γ – Γ) by

regressing E [Π’(k)], or marginal profit of capital, on inside ownership.

2.2 Measuring the Marginal Profit of Capital

From HHL (2002) the marginal profit of capital can be measured as follows.6 Suppose the

firm has a Cobb-Douglas production function Yit = f(Ait;Kit;Zit) = AKαkZαz, where Ait is a measure

of total factor productivity, Yit is output, Kit represents the stock fixed capital including the firm’s

property, plant and equipment, and Zit is a vector of variable factor inputs (e.g., materials, energy,

unskilled production workers). Now assuming that the firm faces an inverse demand curve P(Yit),

variable factor prices wit, and fixed costs F, then the profit function is defined by

Π(Kit;wit) = max P(Yit)Yit –wit Zit – F [4]

s.t. Yit = AKαkZαz

Then by the envelope theorem, the marginal profitability of fixed capital, denoted MPK, is:

6 See also Gilchrist and Himmelberg (1998) for detailed derivation and empirical estimates for US firms.

7

MPK ≡ ∂Πit/∂Kit = (1+ η-1) Pit (∂fit/∂Kit) = (1+ η-1) αk (S/K) = θ (S/K) [5]

Where η is the firm-level elasticity of demand, αk is the capital share of output, S=PY is the

firm’s sales, and θ = (1+η-1)αk is a scale parameter that may vary across industries because price

elasticity of demand as well as the capital share of sales are different across industries. Thus, up to

an industry-specific scale parameter, the ratio of sales to capital may be used to measure the

marginal profitability of fixed capital. Assuming that firms are on average at their equilibrium

capital stocks, the marginal profitability of capital should roughly equal the cost of capital, MPKit =

rit + δit, where rit is the risk-adjusted discount rate and δit is the depreciation rate.

Industry-level estimates of θ can be constructed by averaging over all firms i and years t in

industry j, and by assuming r + δ = 0.18 for all industries. Thus, for industry j, θj is given by:

θj = [(1/NT)ΣiΣt (PitYit/Kit)]-1 (r + δ ) [6]

And, for firm i at time t, Πkit = θj (PitYit/Kit) is the measure of marginal return to capital.

3. The institutional framework and the data

Institutional framework of Italy and Germany

Germany and Italy are generally viewed as countries where investors are weakly protected by

company by-laws. Although they share the Romano-Germanic civil law tradition, Germany

derives its laws from the German Commercial Code, written in 1897 after Bismarck’s unification

of Germany, while Italy’s company laws originates from the French Commercial Code, written

under Napoleon in 1807 (LLSV, 1998). In their 1998’s study, LLSV find several differences

between the two legal families, particularly influencing German’s and Italy’s legal institutions.

While the German law exhibits poor shareholder’s protection, but fairly high protection of

creditors’ rights and law enforcement, Italy’s law extends weak protection to both shareholders and

creditors, as well as medium to low enforcement of law. Finally they both share a low quality of

accounting standards.

8

Introduction of the Neuer Markt segment was a major institutional change for the German

economy (Audretsch and Elston, 2003). The new financial institution had the goal of providing

finance to high technology and high growth startups.7 Admission and reporting standards for Neuer

Markt firms are more stringent than the rules for the regular segment of the Frankfurt exchange.

For example, firms are required to use either International Accounting Standard (IAS) or the US-

GAAP reporting standards in addition to the required Handelsgesetzbuch (HGB) or German

Commercial Code standards, and additionally report in English.

Prior to the European Monetary Union, several institutional features changed in Italy, such as:

liberalization of the financial markets, privatization of many banks and public utilities, and the

introduction of a new corporate law that regulates take-over bids, enforces more informative

accounting standards and disclosure requirements for listed companies. In 1997 the MSE – Milan

Stock Exchange market was privatized, resulting in less restrictive and less costly listing

requirements, while in 1999, the Nuovo Mercato, the Italian counterpart of NASDAQ was launched

for high tech firms –the same year as the Neuer Markt. However in spite of these changes,

investor protection remains comparatively weak in Italy, even according to CONSOB, which

recommends (Annual Report, 2000) that the newly introduced reform of company law (Law n.

58/1998) will “pave the way to strengthened protection of shareholders”. Finally, while the

Corporate Governance Code of Self-discipline was issued in 19998, its enforcement was on a

voluntary basis. Similarly to the US and UK, the main requirements of the Code concern the

nomination of independent, non-executive directors, appointment of the audit, nominating and

compensation committees, and the appointment of an Investor Relations Officer. Although at 2004

almost all listed companies have issued a corporate governance report and made it available on the

Internet, only a minority of the listed companies have adopted at least one of the provisions of the

code.

To conduct our study we constructed two databases, one comprising German new economy

firms on the Neuer Markt, and one comprising Italian manufacturing firms enlisting in the regular

segment of Borsa Italiana.

7 The Neuer Markt segment was introduced by the Deutsche Bourse in 1997 and quickly grew from 2 to 343 firms. Itclosed down in 2002. The firms that remained were placed back on the general Frankfurt exchange – which has sincereorganized into standard and premier segments, with higher reporting standards for the premier listed firms.8 Listed Companies Corporate Governance Committee of Borsa Italiana, “Report and Self-discipline Code”, 1999,available at www.borsaitaliana.it

9

Data

The firm level financial data for the German Neuer Markt comes from three sources -- the

Hoppenstedt database, Deutsche Bundesbank data sources, and publicly available data from the

web, which in total comprise 820 observations (from 1987 to 2001), but not a balanced panel. The

original database includes 341 Neuer Markt firms. Firms included in the database, along with their

initial public offering (IPO) dates, and their industry groupings, are listed in Appendix A. Of these

firms, 13% are not German, but originate from Austria, Britain, France, Israel, Ireland,

Luxembourg, the Netherlands, Switzerland, and the US. The exact number of firms used in

calculating summary statistics and regressions varied somewhat based on data availability for

variables used in that year. Specifically, we excluded firm-year observations with no information

on the identity of shareholders, with sales or fixed capital stock equal to zero. We thus remained

with 160 companies (see Tables A1-A3 in the Appendix for a summary of firm characteristics).

For Italy, we use a panel of Italian manufacturing firms constructed at CERIS-CNR using

several sources.9 Balance-sheet data and IPO characteristics are obtained by Mediobanca, a large

investment bank (the data are collected from two main annual directories: Le Principali Società,

and Indici e Dati); information on corporate governance and ownership structure is derived from

company reports, CONSOB (Commissione Nazionale per le Società e la Borsa, the Authority

which governs Italian equity markets, founded in 1974), and Borsa Italiana (the newly privatized

Italian Stock Exchange). The dataset is an unbalanced panel where firms have at least five

consecutive observations; extensive information about the firm’s income statement, balance sheet

variables is included, along with detailed information about the firms’ age, ultimate ownership,

group-affiliation and business activity. All sample firms are Italian, were initially privately-owned

(non state-owned), conducted an Initial Public Offering in the 1990s and were tracked until 2002.

The earliest entries in the data are in 1995, and the latest entries in 2000, in order to have at least 3

years from the IPO date. To date, the panel comprises 342 firm-year total observations (of which

187 after the IPO) for 43 firms (see Table A2 and A3 in the Appendix), covering the 88% (43 out

of 49) of the total number of manufacturing companies that go public during the sample 1995-2000

period (net of companies that de-listed after just one or two years after the initial offering).

9 For this paper, we extracted the sub-sample of firms going public in the 1990s from a larger dataset (see Benfratello etal., 2001), in which most of the companies are non-public. See also Carpenter and Rondi (2003) for a comparison ofpublic and non-public companies in the 1980s and in the 1990s.

10

In Italy, many quoted firms belong to a large pyramidal business group where at least one

company is publicly traded. In a comparison between firms going public in the 1980s and 1990s,

Carpenter and Rondi (2003) show that all firms listed in the 1990s are independent (non-group

affiliated), family-owned, medium-sized companies. The information on directors’ board

composition in our database confirms prior evidence from LLSV (1999) and Bianchi, Bianco,

Enriques (2001), that controlling families participate in top management, consistently with the

definition of insider in the HHL model.

Inside ownership is a key variable in this study. Ideally inside owners are shareholders that

control decision making (i.e. sitting on the board of directors and/or in the managerial board).10 For

Italian firms we could reconstruct this variable year by year, based on information about boards’

composition and “relevant” shareholdings (i.e. greater than 2% equity fractions) that have to be

notified to CONSOB since 1974 and disclosed to the public since 1995 (Bollettino CONSOB,

Special Issue, various years). This enables us to have a continuous, time-varying (albeit quite

stable) variable. In Italy, the percentage of equity held by the controlling shareholder and by

insiders are very close – 58.7% and 56.0%, on average in our sample, just after the IPO – because

the initial owner/controlling shareholder (the entrepreneur manager who took the company public)

usually sits on the board together with other members of the family. Most often, a «family», or an

individual directly holds the controlling shareholding. Whenever it is held thorough a holding

company, we could identify the ultimate owner in all but two cases.

For German firms, full information about ownership structure and identity of the

shareholders is available at two points in time, prior to and after the IPO. A dummy variable was

constructed to classify the controlling shareholder depending if the firm is controlled (> 50%) by a

family or an individual; non-financial firms; financial service firms; government agencies;

management or disperse; or mixed owner types. Dummy variables representing the concentration

of ownership held among controlling shareholder(s) is also reported in the original source data with

values ranging from 1-4 depending whether the firm is owned by (1-3) individuals with equity

shares greater than 1) 75%, 2) 50%, 3) 25% (each where non one else holds more than 25%), or 4)

a dispersed ownership structure. We used this information in a conservative way, and constructed a

time-invariant dummy variable for “inside ownership” that returns a value of 1 if >50% of the firm

10 HHL (2002) define it as the fraction of equity held by insiders, and use the “closely-held shares” variable from theWorldscope database for their cross-country analysis.

11

is controlled/owned by a family or an individual, or if they own >25% and no one else owns 25%,

and 0 otherwise. Although the lack of a continuous, time varying variable for inside ownership is

somewhat of a limitation in that it prevents us from exploiting the panel dimension, the availability

of ex-ante and ex-post ownership data enables us to investigate what factors determine whether a

firm remains dominated by the same insider (in our case a family or an individual).

4. Empirical strategy and preliminary results

A key feature of the HHL (2002)’s model is that investor protection may vary not only across

countries, but also across firms. “Investor protection is anything that exogenously increases the cost

of stealing from outsiders” (p. 8). This calls for the search of firm-level determinants of investor

protection. For example, tangible assets, such as factories, plants and equipment are difficult to

steal and provide a “built-in” degree of protection to outside shareholders. Or, the identity of

minority shareholders, such as institutional investors or financial institutions, may influence

investor protection to the extent that they carry more (or less) political clout with law enforcement

agencies. Further, in countries with poor investor protection, insiders may have a hard time selling

equity and may wish to signal that they intend to respect their financial contracts. They may do so

by voluntarily disclosing sensitive information, improving quality of accounting standards, or

complying with a self-disciplining code of corporate governance, and so forth.

Hypotheses

The empirical tests in this paper focus on two main hypotheses. First, we test if inside

ownership depends on measures of investor protection by regressing inside ownership on firm-level

variables that proxy the built-in degree of investor protection. Second, we test the prediction of the

HHL (2002) model that the equilibrium level of inside ownership is positively related to the

marginal return of capital, a relationship that reflects the additional idiosyncratic risk premium in

the marginal cost of capital.

To test the first hypothesis for German and Italian firms, we adopt a different estimation

strategy due to the different characteristics of the datasets. We first present the results for the

German firms using a cross-section of firms at the year of the IPO. Table 1 reports the results of a

Logit model estimation in which the probability that the firm is dominated by insiders (as defined

by a controlling family or individual) is a linear and negative function of firm-specific investor

protection. The set of explanatory variables includes firm size (the logarithm of real sales), the sales

12

to capital stock ratio, measuring asset intangibility, a dummy representing a minority shareholding

by a bank or a financial institution, a dummy indicating the presence of venture capital following

the IPO, and the R&D to total asset ratio at the firm level. We also include industry dummies. In

the table we report coefficient estimates for two alternative specifications for the determinants of

inside ownership of German firms –a binomial logit model in columns (i) and (ii), and a

multinomial logit model in columns (iii)-(iv). The first column uses data for the full sample of 160

firms, after excluding firms with zero sales. Columns (ii)-(iv) exclude firms from the financial,

commercial and healthcare services sectors as well as observations where the sale to capital ratio is

less than 0.01.

Our results provide some support of the proposition that inside ownership concentration is

affected by investor protection. Of the variables we propose as determinants of inside

shareholding, venture capital exhibits one of the more significant and negative coefficients in all

specifications, suggesting that the presence of an institutional investor provides some degree of

protection to outside investor. We interpret this as support for the notion that the presence of an

institutional investor assists in facilitating risk diversification by the entrepreneur/manager.

The bank ownership dummy is not significant, which is contrary to our expectation that the

presence of a bank shareholder with better monitoring technology and more political clout with law

enforcement authorities, would serve as a signal of better protection to investors. The sales to

capital ratio is also not statistically significant, but the coefficient is positive, consistent with the

proposition that the greater the proportion of easier-to-divert intangible assets, the weaker the

protection to outsiders against expropriation by insiders and the higher the concentration of inside

ownership. Noticeably, this (weak) evidence surface in a sample of firms operating in industries

where the key assets are characteristically intangible, such as accumulated knowledge,

technological know-how, patents, trademarks, and research and development. In column (ii), where

we include only firms in high-tech sectors, the R&D to total assets ratio turns out to be a more

precise measure for the desire to control for asset intangibility, with a highly significant positive

coefficient.

The two last columns of the table present the results from a multinomial logit analysis where

we extend our definition of inside ownership to include a dummy which returns if an equity share

>50% is held by the management or if no one else own more than 25%. In the three-choice

multinomial logit, the first alternative identifies the insider as the “family” the second alternative

13

identifies the insider as the “management”, with the remaining insiders (financial, non-financial

firms, or government, or mixed) as default. Our results show that refining our typology of insiders

improves the precision of the estimates, at least for family-controlled firms. The sales-to-capital

coefficient is now positive and significant (at the 10% level). The estimated coefficient on venture

capital is negative and highly significant (the p-value is 0.018). The firm level R&D intensity

remains significantly positive at the 6% level.

We now turn to the sample of 43 Italian manufacturing firms listed on Italian Stock Market.

In Table 2, we present coefficient estimates for five alternative specifications for the determinants

of inside ownership concentration, where the dependent variable is the logarithm of α, the relevant

(i.e. > 2%) equity holdings by members of the firm board of directors. The specifications include

several variables of “built-in” investor protection: firm size (the log of real sales), the ratio of sales

to capital stock (where the capital stock is reconstructed at replacement value based on a perpetual

inventory method), a dummy variable which returns if a relevant share of equity is held by an

institutional investor (mutual or pension funds, or venture capital company), a dummy returning if

the firm operates in an R&D intensive industry, and a dummy variable for firms listed in the STAR

(Segmento Titoli con Alti Requisiti) segment of Borsa Italiana. The STAR segment was launched in

2000 by the Italian stock exchange to enhance the visibility of small and medium old-economy

enterprises committing to comply with higher liquidity requirements and more severe disclosure

rules. We then include three dummies indicating if firms have complied with separate Corporate

Governance Code norms (independent directors, audit committee and investor relations) and an

additional dummy which returns a one if the firm has complied with all norms.11 Finally, we

include, mainly for control purposes, a dummy variable that returns a one if a group of the firm’s

shareholders signed an agreement such as a voting pact. As highlighted by Bianchi, Bianco and

Enriques (2001), in Italy, voting pacts and other shareholders agreements are viewed as devices to

separate control from ownership as they allow insiders to control the firm with small equity

stakes.12

11 The collection of this variable required special attention for two reasons. First, companies are only recommended, notrequired complying with the code. Second, many companies provided only vague descriptions of their compliance tothe code (e.g. not indicating the number or names of independent directors, or whether non-executive and independentdirectors in the required proportions were in the auditing committee). Only companies that applied all three normsliterally were assigned a dummy value of 1.12 Listed firms have to notify shareholders’ agreements, their content and duration to CONSOB.

14

In column (i) we report the results from an OLS regression which add industry and year

dummies. Firm size enters with a positive and significant coefficient, indicating that inside

ownership is higher at larger firms. This is in contrast with HHL’s finding of a negative coefficient,

which they explain by arguing that large size may ensure better protection to outside investors

because of economies to scale to monitoring. Our finding, however, appears more consistent with

the Italian institutional context, where small and medium firms appear motivated to adopt

monitoring devices by the need to raise external finance in a weakly protected environment. We

also find a highly significant negative coefficient on the dummy variable indicating “Shareholders’

Agreement”. As many of the sample firms have shareholders that signed voting pacts, failure to

control for this would distort our estimates, because insiders de facto control the firm with low

equity stakes. Of the remaining variables, proxying for the built-in degree of investor protection –

R&D intensity, STAR segment, corporate governance compliance – only the Institutional Investors

dummy enters with the predicted negative sign, similar to German firms.

Coefficients are more precisely estimated when we also control for firm-specific effects in

columns (ii)-(v) using Least Squares Dummy Variables regressions. The coefficient on the Sales to

capital ratio is positive and not far from significance in col. (ii) and (iii). As long as low ratios

indicate asset tangibility, i.e. assets which are difficult to divert or steal, this would indicate

stronger built-in protection to outside investors, and would make it easier for the insider to sell a

larger share of the firm’s equity. By reflection, and consistently with the model, intangibility of the

assets (e.g. technological and marketing capabilities, R&D, etc.) predicts higher levels of inside

equity.

The coefficient on the Institutional Investors dummy is negative, but not significant. The

dummy for firms that enlist in the STAR segment is also insignificant. However, when we enter

them interactively, they turn significantly negative, as predicted by the model, but only the latter is

robust to the introduction of the time dummies.

The Corporate Governance dummy is the variable that we expect to best capture the built-in

degree of investor protection. Consistently with the model predictions, we find that the estimated

coefficient is negative and significant in all specifications. When we enter the code’s dummies

separately we find that both the presence of Independent Directors and the existence of the Audit

Committee display the expected negative effect on inside equity ownership.

15

The second test we carry out for Italian firms focuses on the predicted positive relationship

between the level of inside ownership and the marginal return to capital.13 In Table 3, we report the

estimated coefficients from simple LSDV regressions of the marginal profitability to capital (MPK)

on (lagged) inside ownership concentration for our sample of Italian firms (preliminary results):

MPKit+1 = r + δ + Γ + (γ – Γ)α it +uit [7]

Following HHL (2002) model, these regressions produce estimates of (γ – Γ), which reflects

the average additional risk premium for bearing idiosyncratic risk, above the systematic risk

premium, which is absorbed in the constant term. The results in columns (i) and (ii) show that the

inside ownership coefficients are positive and significant, with estimated values of 0.072, and of

0.054 when we add the year dummies.

This result, though preliminary, is consistent with HHL’s model prediction that, under poor

investor protection, the higher is the concentration of inside ownership, the higher is the implied

cost of capital. This suggests that the magnitude of the distortions from the first-best level of capital

stock may be large in a country like Italy, thus concurring to explain underinvestment and slow

growth.

5. Conclusions

In this paper, we empirically investigate the impact of investor protection on the cost of

capital in Europe, where investor protection characterizes what Himmelberg, Hubbard and Love

(2002) collectively refer to as “those features of the legal, institutional, and regulatory environment

– and characteristics of firms or projects - that facilitate financial contracting between inside

owners (managers) and outside investors.”

To date, the broad international studies within the Law and Finance literature indicate the

significance of these effects overall, and the importance of examining differences between

countries. Our study extends previous work by 1) providing evidence of firm behavior and agency

related problems for young firms in the 1990s in Germany and Italy, and 2) allowing for distinction

13 In the Appendix, Table A4, we report the two-digit NACE industry estimates of the scale parameter θj that we usedto construct the firm-level marginal profitability of capital variables. Our estimates are based on the full sample of over1700 firms included in the CERIS dataset. We also include the industry estimates constructed by Gilchrist andHimmelberg (1998) for comparison.

16

between degrees of “poor shareholder protection” countries as nuanced between Germany and Italy

which have two very different institutional environments.

Again, in countries where investor protection is said to be generally low, the HHL model

predicts endogenously high levels of insider ownership. Accordingly, the idiosyncratic risk

premium applied to the cost of capital should be high, implying a steady-state level of capital below

the first-best level.

Using firm level data from both Italy and Germany, we estimate the determinants of the

fraction of equity owned by insiders, testing, as predicted, that this fraction depends on measures of

investor protection. We also test the HHL prediction that investor protection has an important

cross-firm and cross-cultural dimension. In addition, we investigate the correlation between inside

equity ownership and the marginal return to capital, a relationship that follows directly from the

first-order condition for capital –the cost of capital includes a risk premium that reflects the

insiders’ exposure to idiosyncratic risk.

Our findings suggest that for Italy voluntary compliance with Corporate Governance norms

or tighter liquidity and disclosure rules has a negative impact on the concentration of inside

ownership of the firm, while in Germany venture capital ownership had a similar negative impact

on inside ownership concentration. Unlike Italy, a German firm’s R&D intensity had a strongly

positive impact on the probability that the firm would have concentrated insider ownership,

indicative that R&D firms may have a hard time obtaining equity capital. Consistent with these

results is the fact that in Germany bank ownership had no impact on ownership concentration

which is not surprising given the minimal involvement of banks with new economy firms in

Germany. In Italy, institutional investment is a negative determinant of inside ownership, but only

when coupled with more stringent disclosure and corporate governance rules. In both countries,

asset intangibility is a positive determinant of inside ownership.

We also find preliminary evidence that in Italy, the higher the concentration of inside

ownership, the higher is the implied cost of capital. This suggests that the magnitude of the

distortions from the first-best level of capital stock may be large in countries with poor investor

protection, like Italy and Germany, concurring to explain underinvestment and slow growth.

17

References

Alchian, A., Demsetz, H., 1972. Production, Information Costs, and Economic Organization.American Economic Review 62: 7767-805.

Audretsch, D.B., Elston, J.A. (2003), Can Institutional Change Impact High-Technology FirmGrowth? Evidence from Germany’s Neuer Markt. Unpublished working paper, Universityof Central Florida, May.

Barca, F., Becht, M. (Eds.), 2001. The Control of Corporate Europe. Oxford University Press.

Benfratello, L. et Al. (2001), Il nuovo panel Ceris su dati di impresa. Working Paper CERIS CNR,N. 5, Torino, October.

Bianchi, M., Bianco, M., Enriques, L., 2001. Pyramidal Groups and the Separation BetweenOwnership and Control in Italy. In: Barca F. and M. Becht, (eds.), The Control of CorporateEurope. Oxford University Press.

Carpenter, R.E., Rondi L., 2003. The Growth and Performance of Newly Public Italian Firms.Unpublished working paper, Ceris-Cnr, July.

Demirgüc-Kunt, A., Maksimovic, V., 1998. Law, Finance, and Firm Growth. Journal of Finance53: 2107-2137.

Gilchrist, S., Himmelberg, C.P., 1998. Investment, Fundamentals, and Finance. In: Ben Bernankeand Julio J. Rotemberg (eds.), NBER Macroeconomics Annual. MIT Press, Cambridge.

Himmelberg, C.P., Hubbard, R.G., Love, I., 2002. Investor Protection, Ownership, and the Cost ofCapital. Unpublished working paper, Columbia University, April.

Hubbard, R.G., 1998. Capital-Market Imperfections and Investment. Journal of EconomicLiterature 36, 193-225.

Jensen, M., Meckling, W., 1976. Theory of the Firm: Managerial Behaviour, Agency Costs, andOwnership Structure. Journal of Financial Economics 3, 305-360.

La Porta, R., Lopez-de-Silanes, F., Shleifer, A., Vishny, R., 1998. Law and Finance. Journal ofPolitical Economy 106, 1113-1155.

La Porta, R., Lopez-de-Silanes, F., Shleifer, A., Vishny, R., 1999. Ownership Around the World.Journal of Finance 54, 471-517.

La Porta, R., Lopez-de-Silanes, F., Shleifer, A., Vishny, R., 2000. Investor Protection andCorporate Governance. Journal of Financial Economics 58, 3-27.

Levine, R., 1999. Law, Finance, and Economic Growth. Journal of Financial Intermediation 8, 8-35.

Pagano, M., Panetta, F., Zingales, L. (1998), Why Do Companies Go Public. An EmpiricalAnalysis. Journal of Finance, LIII, 1, 425-448.

Rajan, R., Zingales, L., 1998. Financial Dependence and Growth. American Economic Review 88,559-586

18

TABLE 1LOGIT ANALYSIS OF THE PROBABILITY OF INSIDE OWNERSHIP IN GERMAN NM FIRMS

(t-statistics in parentheses)

Binomial Logit Models Multinomial LogitFamilyOwned

FamilyOwned

FamilyOwned

Manage-ment Owned

Inside Ownership

defined as: (i) (ii) (iii) (iv)

Log (Sales) 0.072 0.053 -0.127 -0.207 (0.627) (0.303) (-0.551) (-0.901)

Sales / Capital 1.352 1.891 4.243* 3.072 (1.317) (1.281) (1.63) (1.16)

Bank Ownership 0.126 -0.256 0.486 1.013 (0.289) (-0.475) (0.665) (1.457)

Venture Capital -0.730* -0.823* -1.461** -1.359** (-1.742) (-1.630) (-2.485) (-2.357)

R&D / Total Assets 2.508 12.467** 17.140* 8.579 (1.113) (2.242) (1.887) (0.918)

Industry Dummies Yes Yes No

χ2 15.6 [14] 15.5 [11] 17.5 [10]p-value 0.34 0.16 0.06

N. Firms 160 118 118

Note:*** Coefficient significantly different from 0 at the 1 percent level or less.** Coefficient significantly different from 0 at the 5 percent level.* Coefficient significantly different from 0 at the 10 percent.

19

TABLE 2DETERMINANTS OF INSIDE OWNERSHIP CONCENTRATION FOR ITALIAN LISTED FIRMS

(t-statistics in parentheses)

Dependent Variable = Log α

LSDVFirm-Level CharacteristicsOLS

(i) (ii) (iii) (iv)

Log (Sales) 0.043* 0.081* 0.092* 0.101 0.116(1.725) (1.627) (1.855) (1.365) (1.562)

Sales / Capital 0.003 0.006 0.007 0.004 0.004(0.429) (1.520) (1.577) (0.876) (0.856)

R&D Intensity Dummies (3-digit) -0.037 - - - -(-0.666) - - - -

STAR Segment 0.027 -0.017 -0.128*** -0.111* -0.072(0.598) (-0.475) (-2.829) (-1.826) (-1.128)

Institutional Investors -0.106** -0.014 -0.063*** -0.058 -0.045(-2.309) (-0.636) (-2.694) (-1.533) (-1.184)

STAR * Inst. Inv. - - 0.113*** 0.107** 0.101*- - (2.991) (2.057) (1.848)

Corporate Governance Code 0.005 -0.121* -0.127* -0.118* -(0.061) (-1.691) (-1.784) (-1.658) -

Independent Directors - - - - -0.074- - - - (-1.439)

Audit Committee - - - - -0.283***- - - - (-2.992)

Investor Relator - - - - 0.168*- - - - (1.811)

Shareholders’ Agreement -0.129*** 0.024 0.025 0.022 0.043(-2.761) (0.420) (0.448) (0.432) (0.777)

Year Dummies Yes No No Yes Yes2-digit Industry Dummies Yes No No No No

Adj-R2 0.229 0.658 0.659 0.647 0.679N. obs. 187 187 187 187 187N. Firms 43 43 43 43 43

Note: Standard errors corrected for heteroskedasticity*** Coefficient significantly different from 0 at the 1 percent level or less.** Coefficient significantly different from 0 at the 5 percent level.* Coefficient significantly different from 0 at the 10 percent.

20

TABLE 3ESTIMATES OF THE FIRST-ORDER CONDITION FOR THE CAPITAL STOCK – ITALIAN SAMPLE

(t-statistics in parentheses)

Dependent Variable = MPK

LSDV

(i) (ii)

Inside Ownershipt-1 0.072*** 0.054**(4.441) (2.041)

-Year 1997 - 0.003

- (0.797)

Year 1998 - 0.012***-- (2.656)

Year 1999 - 0.005- (1.250)

Year 2000 - 0.007- (0.895)

Year 2001 - -0.011- (-1.260)

Year 2002 - -0.023**- (-2.371)

Year Dummies No Yes

Adj-R2 0.984 0.984N. obs. 144 144N. Firms 43 43

Notes: Standard errors corrected for heteroskedasticity.*** Coefficient significantly different from 0 at the 1 percent level or less.** Coefficient significantly different from 0 at the 5 percent level.* Coefficient significantly different from 0 at the 10 percent.

21

Appendix

Table A1 – Firm characteristics for the sample of German listed companies.

Industries N. Firms % of Inside-Controlled

Firms

% of Firmswith Venture

Capital

% of Firmswith Bank

shareholding

1 Biotechnology 6 17 17 172 Financial Services 1 0 0 03 Commercial Services 11 45 18 184 Internet 22 41 50 185 Software 20 55 25 156 Media and Advertising 20 50 30 307 Healthcare 7 43 57 298 Data Processing 27 30 19 199 Electronics 38 34 32 24

10 Telecommunications 12 25 33 25

Table A2 - Firm characteristics for the sample of Italian listed companies

Industries N.Firms

Insideequity

ownership

STARsegment

InstitutionalInvestor

Shareholders’Agreement

CorporateGovernance

1 Non metallic mineral products 2 55.95 0.00 1.00 0.50 0.002 Chemical rubber and plastics 4 58.87 0.50 0.75 0.50 0.003 Machinery and Equipment 8 46.87 0.38 0.88 0.50 0.25

4 Electrical machinery, TLC andelectronics 13 60.31 0.46 0.38 0.31 0.23

5 Transport Equipment 4 50.90 0.50 0.75 0.25 0.756 Food and Drinks 4 52.68 0.75 0.25 0.25 0.507 Textile and Clothing 7 59.40 0.29 0.14 0.29 0.298 Print Publishing 1 70.56 0.00 0.00 0.00 1.00

Table A3 - Timing of the Initial Public Offerings

German firms Italian firms

1992 1 01995 1 81996 2 51997 8 71998 32 61999 75 92000 46 8

165 43

22

Table A4 - Two-Digit NACE estimates of θj for the CERIS Panel of Italian Firms

S/KNACE_Rev0 Obs.

Mean Std. Dev.

θJ (SK)ITA SIC θJ SK1

G-H

24 1007 1.349 0.785 0.133 32 0.06925 1946 3.679 3.723 0.049 28 0.05132 1667 3.919 4.007 0.046 35 0.03633 69 2.199 1.457 0.082 35 0.03634 1788 4.261 5.350 0.042 36 0.03935 434 3.895 6.564 0.046 37 0.03736 428 2.791 3.278 0.064 37 0.03737 246 4.208 4.782 0.043 38 0.03641 1188 5.262 4.956 0.034 20 0.03643 1067 2.655 2.685 0.068 22 0.03545 707 5.446 5.032 0.033 23 0.01747 1066 3.195 3.229 0.056 26-27 0.05748 545 2.667 3.874 0.067 30 0.04049 151 6.450 6.931 0.028 39 0.032

Corr ITA-GH 0.757

Note: The last column reproduces 2-digit SIC estimates of θj by Gilchrist & Himmelberg (1998).